Many cancer-associated mutations remain uninvestigated because experimental validation of their effect on gene function is costly, time-consuming, and requires prior knowledge of a gene's function. In this dissertation, I present the expression-based Variant Impact Phenotyping (eVIP2) approach, which uses gene expression data to characterize a gene variant's function, requiring no prior knowledge of the wild-type gene's function. With eVIP2, we can predict if a mutation causes a gain, loss, or change in function, or if it is neutral. We determined that two recurrent frameshift mutations in RNF43 have different effects on gene function where one mutation causes multiple cancer pathways to be activated.

Next, I explore the role of alternative splicing in multiple myeloma therapy. Unbiased phosphoproteomics revealed that phosphorylation events on splicing factors are upregulated after treatment with a proteasome inhibitor (PI), which led us to explore the spliceosome as a multiple myeloma vulnerability. We found PI treatment leads to prominent intron retention, suggesting splicing interference as an unrecognized modality of PI mechanism.

Finally, I present sigil, a tool for parallel creation of representative gene expression and alternative splicing signatures. Microarray-based immunological gene expression signatures are widely used for gene set enrichment analysis and better understanding of disease. However, despite evidence of splicing differences across cell lineages, there is no equivalent resource for alternative splicing. Here, we use sigil to create RNA-seq-based gene expression and splicing signatures which are compatible with enrichment analysis. We identify prevalent alternative promoter usage and enrichment for cell-type specific junctions associated with differential inclusion of protein modifications.